The inference speed is slower than yolov3

[blackwool]

I test this repo using 416*416 size. under win10 , NVIDIA GTX 1060 GPU
Each picture inference time is about average 112ms.
But when i use stand yolov3.cfg to detector ordinary object ,each picture takes about average 38 ms .
Why this network is so much slower than yolov3 network ?

[Bardielz]

hi，in the issue 'the test speed',the author says under 2080ti input size:608x608 ,the image ./test_imgs/input/selfie.jpg: Predicted in 6.259000 milli-seconds. may be in the makefile the folowing selects such as "GPU=1" you don't choose ,so the speed you test is worked under cpu?that is my guess ,may not be true,but you can have a try.

[blackwool]

I do use GPU=1 configure to make file ,and it calls cudnn api when it runs.
I print each layer costed time and find the that the convolutional layers with large groups will cost a lot of time .

[w840401]

How did you train successfully?